If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The prevalence of diabetes is rapidly increasing worldwide. Type 2 diabetes may remain undetected for many years, leading to severe complications and healthcare costs. This paper provides estimates of the prevalence of undiagnosed diabetes mellitus (UDM), using available data from high quality representative population-based sources.
Methods
Data sources reporting both diagnosed and previously undiagnosed diabetes were identified and selected according to previously described IDF methodology for diabetes in adults (aged 20–79). Countries were divided into 15 data regions based on their geographic IDF Region and World Bank income classification. The median UDM proportion was calculated from selected data sources for each of data region. The number of UDM cases in 2013 was calculated from country, age and sex-specific estimates of known diabetes cases and data region-specific UDM proportion.
Results
Of 744 reviewed data sources, 88 sources representing 74 countries had sufficient information and were selected for generation of estimates of UDM. Globally, 45.8%, or 174.8 million of all diabetes cases in adults are estimated to be undiagnosed, ranging from 24.1% to 75.1% across data regions. An estimated 83.8% of all cases of UDM are in low- and middle-income countries. At a country level, Pacific Island nations have the highest prevalence of UDM.
Conclusions
There is a high proportion of UDM globally, and especially in developing countries. Further high-quality studies of UDM are needed to strengthen future estimates.
]. The International Diabetes Federation (IDF) estimates that there are 381.8 million people with diabetes in 2013 with a projected increase of 55% to 591.9 million by 2035 [
]. As a result of a combination a number of factors including: under-performing health systems, low awareness among the general public and health professionals, and the often slow onset of symptoms or progression of type 2 diabetes, the condition may remain undetected for many years, during which time complications may develop. Population-based studies actively screening for diabetes using either oral glucose tolerance test (OGTT) or fasting blood glucose provide the backbone for estimating undiagnosed diabetes (UDM). In such studies, participants who report not having been diagnosed with diabetes may be found to have diabetes upon testing of their blood glucose and would therefore be classified as having UDM, i.e., ‘previously undiagnosed’ or ‘newly diagnosed’ diabetes. While it is possible to have undiagnosed type 1 diabetes, this is usually short in duration due to the rapid onset of symptoms, and would not likely be measured in the population-based studies necessary for the estimation of undiagnosed diabetes. However, few studies reporting the prevalence of diabetes make a distinction between type 1 and type 2 diabetes, and it is therefore not possible to separate any estimate of undiagnosed diabetes.
The prolonged asymptomatic phase of type 2 diabetes may last many years [
], during which time unmanaged elevated blood glucose leads to serious and irreversible development of micro- and macrovascular complications including neuropathy, nephropathy, retinopathy, coronary artery disease, stroke and peripheral vascular disease [
]. Rates of complications have been shown to be high in people with UDM compared to normoglycaemic individuals. In the USA, up to 41.7% of adults with previously undiagnosed diabetes have chronic kidney disease [
]. Furthermore, BMI, blood pressure, and other cardiovascular and metabolic markers have been found to be significantly higher in a cohort of people with coronary artery disease and also UDM compared with diagnosed diabetes; likely due to awareness of the condition and subsequent dietary modifications [
]. Undiagnosed diabetes has been reported to carry a similar risk of mortality to diagnosed diabetes, and is associated with a 1.5- to 3.0-fold higher risk of mortality compared to normoglycaemic individuals [
Without the mechanisms and resources necessary for early detection, a person with diabetes may only be diagnosed after the onset of complications. Prevention of complications in people with diabetes by timely lifestyle and pharmaceutical interventions has been shown to reduce hyperglycaemia and risk of complications, [
]. In addition to a heavy health burden, the financial costs of diabetes-related health expenditures weigh heavily on individuals, health systems and governments, with global health expenditure estimated to be at least 548.5 billion USD in 2013 [
]. The cost of undiagnosed diabetes may contribute substantially to this estimate. One study from the USA found that an additional 2864 USD were spent on direct and indirect costs per person with UDM per year, or 18 billion USD nationally [
]. While the cost of screening for and subsequently treating diabetes is considerable, it is far outweighed by the cost of treating potentially preventable diabetes-related complications [
Estimates of direct medical costs for microvascular and macrovascular complications resulting from type 2 diabetes mellitus in the United States in 2000.
]. It is important to produce regional and global estimates of UDM in order to understand the burden of UDM globally and regionally, its drivers and potential implications for policy and practice.
While the existence of UDM has long been recognised [
], wide-reaching awareness among the general public, physicians and policy-makers is lacking and there are limited reliable and comparable data available on the subject. Nationally representative population-based studies using OGTT are considered the gold standard for studying the prevalence of diabetes and quantifying undiagnosed diabetes [
]. However, the availability of these studies is varied and may be limited for similar reasons to those which cause diabetes to go undiagnosed; namely that screening effectively for diabetes is costly, time consuming, and, for many countries, not a priority. However these considerations should be balanced by the costs and health burden associated with UDM.
], providing a global-scale quantification of this burden. An accurate estimation of the burden of UDM is highly relevant given the high health-related and financial costs associated with diabetes.
Given the lack of awareness and considerable burden of UDM, this paper presents a standardised method and accompanying results for country, regional and global estimates of UDM for the year 2013. These estimates are included in the 6th edition of the IDF Diabetes Atlas [
] and is summarised in Fig. 1. Briefly, data sources reporting the prevalence of diabetes were identified through a systematic literature search for the period November 2010–June 2013, using PubMed, Google Scholar, websites of governments, the World Health Organization and associated organisations, personal communication with investigators in the IDF network, and by searching reference lists.
Fig. 1Study and data source selection and generation of estimates for undiagnosed diabetes mellitus (UDM) in adults (20–79 years), 2013.
Data from 744 retrieved data sources were stored in database MySQL database and characterised by the following criteria: type of data (e.g. peer-reviewed publication), study design (e.g. population-based), sample representation (e.g. regional representation), diagnostic criteria (e.g. oral glucose tolerance test – OGTT), sample size, and study year. Data sources lacking sufficient data on age-specific prevalence of diabetes, essential information on methodology, or details on study characteristics were excluded, and also duplicate or out-dated data, and data from clinic- or hospital-based studies. The study characteristics were then used to score each data source using a scoring system based on the Analytic Hierarchy Process [
]. Data sources that were nationally representative, population-based, used oral glucose tolerance test (OGTT), and were conducted in the last five years were favoured. Sources including data on both self-reported (i.e., diagnosed prior to the study) and previously undiagnosed (i.e., diabetes first diagnosed during the study) were used for the estimation of UDM.
2.2 Aggregation and calculation of proportion of UDM
The proportion of UDM (i.e. percentage of all cases of diabetes that are undiagnosed) was extracted or calculated from data sources reporting both diagnosed and previously undiagnosed diabetes. Due to limited data availability, countries were grouped into 15 data regions grouped by a combination of the seven IDF Regions (Africa – AFR; Europe – EUR; Middle East and North Africa – MENA; North America and the Caribbean – NAC; South and Central America – SACA; South East Asia – SEA; and Western Pacific – WP) and World Bank Income classification group (high, middle and low income countries; HIC, MIC and LIC, respectively [
] to produce a combined category, or “data region” (e.g. AFR-MIC). No data sources were available for MENA-LIC or SACA-HIC. Therefore, countries which were in the data regions MENA-LIC (namely, Afghanistan) and SACA-HIC (namely, Puerto Rico) were grouped with MENA-MIC and NAC-HIC, respectively. Sets of source data on the proportion of UDM were grouped by data region and the median was calculated to arrive at the estimated proportion of UDM for each data region.
2.3 Estimates of cases and prevalence of undiagnosed diabetes
The total number of cases in adults (aged 20–79) for 219 countries and territories were calculated using the estimated UDM proportion matched by data region and the country-, sex-, and age-specific estimates of the number of adults with diabetes. Cases of UDM were aggregated to produce global and regional estimates of UDM, as well as estimates by income group. The prevalence of UDM (i.e. the percentage of the population with UDM) was calculated by dividing the age-specific number of cases of UDM in adults by the age-specific adult (20–79 years) population [
United Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects: The 2012 Revision [Internet]. New York; 2013. Available from: http://esa.un.org/wpp/Documentation/publications.htm.
Of the 744 reviewed data sources, 174 were used for diabetes prevalence estimates in adults (20–79 years); 88 sources with information on known and previously undiagnosed diabetes were selected (Fig. 1, Appendix).
Table 1 presents the UDM proportion by data region, including study characteristics and range of data that contributed to the regional estimate. Overall, data were drawn from 74 of 219 countries (33.8%). Middle-income countries had the highest proportion of countries represented by original source data (39.4%; n = 43) compared to HICs (28.4%; n = 19) and LICs (27.9% n = 12) countries. By IDF Region, SEA had the highest proportion of countries with selected data (71.4%) followed by SACA (55.0%), and MENA (50.0%). Only 25.0% of countries in AFR had data included in the estimates. Just over half of all 88 selected data sources were based on OGTT (54.2% in HICs, 53.8% of studies in MICs, and 33.3% in LICs, respectively), and 75% of all data sources were nationally representative (95.8% of studies in HICs, 67.3% in MICs and 66.7% in LICs, respectively).
Table 1Study sources, characteristics by data region and proportion of undiagnosed diabetes (UDM) in adults (20–79 years), 2013.
Data region
No. of studies
Countries in data region providing data
Countries (no. of studies)
Study sample representation
Study diagnostic
Study range DM prevalence (%)
Study range UDM proportion (%)
Data region UDM proportion (%)
AFR-MIC
5
4/17; 23.5%
Angola (1); Réunion (1); Seychelles (1); South Africa (2)
es (1); nat (2); reg (2)
OGTT (5)
2.8–20.1
35.5–91.7
46.0
AFR-LIC
8
8/31; 25.8%
Benin (1); Comoros (1); Guinea (1); Kenya (1); Mozambique (1); Niger (1); Togo (1); United Republic of Tanzania (1)
nat (7); reg (1)
FBG (6); OGTT (2)
2.6–10.5
36–99.1
75.1
EUR-HIC
11
10/33; 30.3%
Croatia (1); Cyprus (1); Estonia (1); Finland (1); France (2); Hungary (1); Malta (1); Portugal (1); Spain (1); United Kingdom (1)
loc (1); nat (10)
FBG (5); OGTT (6)
5.1–14.6
16.7–60.9
36.6
EUR-MIC
4
4/20; 20.0%
Albania (1); Bulgaria (1); Poland (1); Turkey (1)
nat (3); reg (1)
FBG (2); OGTT (2)
4.2–16.5
27.7–45.5
35.1
EUR-LIC
1
1/3; 33.3%
Uzbekistan (1)
reg (1)
OGTT (1)
8.2
29.3
29.3
MENA-HIC
5
3/6; 50.0%
Oman (2); Saudi Arabia (2); United Arab Emirates (1)
nat (5)
FBG (3); OGTT (2)
10.2–23.7
27.9–50
40.7
MENA-MIC
12
8/15; 53.3%
Algeria (2); Egypt (1); Iraq (1); Islamic Republic of Iran (1); Jordan (3); Pakistan (1); State of Palestine (2); Tunisia (1)
nat (11); reg (1)
FBG (7); OGTT (5)
5.9–25.9
21.7–75
50.0
NAC-HIC
3
3/14; 21.4%
Barbados (1); United States of America (1); US Virgin Islands (1)
nat (3)
HbA1c (1); OGTT (2)
13.7–17.5
10.2–39.8
27.7
NAC-MIC
3
3/12; 25.0%
Belize (1); Guadeloupe (1); Jamaica (1)
nat (3)
FBG (2); OGTT (1)
6.6–13.1
18.5–41.2
25.0
NAC-LIC
1
1/1; 100%
Haiti (1)
reg (1)
OGTT (1)
11.3
29.4
29.4
SACA-MIC
11
11/19; 57.9%
Argentina (1); Bolivia (1); Chile (1); Colombia (1); Costa Rica (1); Ecuador (1); Guatemala (1); Honduras (1); Nicaragua (1); Peru (1); Venezuela (1)
loc (2); nat (1); reg (8)
FBG (6); OGTT (5)
4.4–9.4
20.0–48.8
24.1
SEA-MIC
5
4/5; 80.0%
Bhutan (1); India (1); Mauritius (2); Sri Lanka (1)
loc (1); nat (4)
OGTT (5)
8.2–17.9
35.8–69.5
49.1
SEA-LIC
1
1/2; 50.0%
Nepal (1)
reg (1)
FBG (1)
14.6
43.6
43.6
WP-HIC
5
3/13; 23.1%
Hong Kong SAR (1); Republic of Korea (2); Singapore (2)
nat (5)
FBG (2); OGTT (3)
7.6–9.5
31.9–64.4
49.4
WP-MIC
12
9/21; 42.9%
China (2); Fiji (1); Indonesia (1); Malaysia (3); Mongolia (1); Nauru (1); Samoa (1); Thailand (1); Tonga (1)
nat (11); reg (1)
FBG (7); OGTT (5)
5.7–16.0
27.4–78.5
54.1
WP-LIC
1
1/5; 20.0%
Cambodia (1)
nat (1)
FBG (1)
2.9
63
63.0
Es, ethnic or specific group; loc, local; nat, national; reg, regional; FBG, fasting blood glucose; OGTT, oral glucose tolerance test.
Globally, 45.8% of all cases, or 174.8 million people, are estimated to have UDM in 2013. There is considerable variability in the UDM proportion across different regions, ranging from 24.1% in SACA-MICs to 75.1% in AFR-LICs (Table 1). Disaggregation of MICs to examine upper middle income countries (UMICs) and lower middle income countries (LMICs) alongside HICs and LICs shows that LMICs have the highest number of cases of UDM of all income groups, with 108.4 million. Overall, 83.8% of all cases of UDM are in LICs and MICs (Fig. 2). At the country level, Tokelau is estimated to have the highest prevalence of UDM, at 20.5%, followed by the Marshall Islands at 18.9% and the Federated States of Micronesia at 16.1%, whilst Azerbaijan has the lowest at 0.8% (Table 2). The 10 countries with the highest UDM prevalence are all Pacific Islands.
Fig. 2Total number of cases (1000s) of diabetes, both undiagnosed (UDM) and diagnosed, in adults (20–79 years), 2013, by World Bank income group [
Table 2Countries/territories with highest and lowest prevalence (%) and number of cases (1000s) of undiagnosed diabetes mellitus (UDM) in adults (20–79 years), 2013.
Globally, 174.8 million people are estimated to have UDM using the standardised IDF methodology described in this paper. These estimates confirm that lack of detection of diabetes persists throughout the world, across all regions and income groups. Despite a variation in availability and quality of data, studies were available for every IDF Region and all income groups (Table 1).
While there is wide variability in the proportions of UDM from the underlying sources for the AFR-LIC region (Table 1) with a study from Togo reporting a proportion of 99.1% [
], as a whole, this data region has the highest proportion of UDM (75.1%). Not coincidentally, these countries represent some of the least developed health systems in the world and have been focused on a large infectious disease burden which likely contributes to a low awareness of NCDs and diabetes. Although the next highest proportion of UDM is found in WP-LIC countries, it is the middle-income countries in that region which contribute a large number of cases to the UDM burden. The WP-MIC statistic is based largely on estimates for China [
] with an estimated 54.1% of diabetes cases undiagnosed. However, new evidence shows that this may be an underestimate as a recent nationally representative survey of Chinese adults reported a proportion of UDM of 70% [
NAC-MIC and SACA-MIC have very similar UDM proportions (25.0% and 24.1%, respectively), reflecting the comparable states of health system development and state of the diabetes epidemic in those regions. Conversely, the estimate for NAC-LIC, which includes only Haiti, is based on a study from the capital city and is likely an underestimate of the national proportion of UDM which would include rural areas with a lower access to healthcare and likely poorer awareness of diabetes. For example, a study from Guinea, a country with a similar level of development to Haiti, reported a UDM proportion of 100% in the rural population, compared to 58.8% in the urban population [
]. The greatest range of data comes from AFR-LIC (36.0–99.1% across 8 studies), reflecting the broad range of GNI per capita within the LIC income group. Where more studies are available, the estimates may benefit from a narrower grouping of countries matched more closely for health system development.
The countries with the highest prevalence and number of cases of UDM (Table 2) closely follow those for total diabetes. Seven of the top 10 countries for prevalence of UDM also feature in the top 10 countries for total diabetes prevalence, while this applies to 9 of the top 10 countries for cases. LMICs have by far the greatest number of cases of UDM, followed by HICs (Fig. 2). There is a complex interplay of factors contributing to UDM, rather than UDM being directly related to income. These determinants may include established risk factors associated with diabetes, but also awareness of the condition and access to and quality of healthcare [
]. Health systems are generally more developed in HICs (in terms of number, education, and resources available to health professionals) and thus diabetes may be detected earlier than in less developed regions where more barriers exist. In addition healthcare access may be lower in rural areas within countries in any income region [
]. These factors exert varying degrees of influence in different countries within the income regions concerned, but may account for the variation. The overarching truth is that UDM will continue to be a public health issue even as economic development continues, unless commitments are made to strengthening health systems, with a particular focus on resource-poor regions.
Data availability for the estimates varied considerably by region and especially by income group. Globally, 33.8% of countries from all seven IDF Regions were represented (Table 1). A higher proportion of MICs (39.4%) was represented than HICs and LICs (28.4% and 27.9%, respectively). High-income countries had by far the highest proportion of nationally representative data sources; however, there are a greater proportion of large nationally representative studies based on self-reported data in HICs with less focus on OGTT-based screening compared to MICs. This further illustrates that availability of resources does not necessarily correlate with a wealth of usable data for UDM.
5. Limitations
The principal limitation in generating accurate estimates for UDM is the lack of high quality data suitable for inclusion. Given that the data regions with sparse data sources tend to be those comprised of LICs and MICs, consideration should be given to how these countries could be supported. Many countries invest in large scale studies, but base these on self-reported data, or may either not record whether subjects had previously been diagnosed with diabetes before screening, or not report these data. A standardised approach to conducting studies and presenting results, with additional recommendations for developed settings would be beneficial. Issues pertaining to variability in methodology and suitable standardisation methods could also be resolved by a set of stepped study guideline recommendations.
While the regional and global estimates in this report accurately reflect the available data, there is considerable intra-region and intra-country variation which is not reflected. The number of data sources is further limited by the selection criteria: while 292 of the 744 sets of data in the database contained data on UDM, only 88 (those which were also selected for generating estimates of diabetes prevalence) were used to generate estimates for UDM. Future estimates may benefit from a separate set of selection criteria for papers used to generate the estimates for UDM independent of the criteria for selection for overall prevalence estimates. Sensitivity analyses are underway to determine the effect of modifying the inclusion criteria for UDM on the global prevalence estimates.
A further limitation is the lack of information on age-specific proportions of UDM. Evidence from six data sources (Fig. 3) show that proportions across different regions tend to be higher for younger age groups and decrease with age [
]. The cause of the higher UDM proportions in younger age groups is likely two-fold. Firstly, routine screening tends to target older individuals, for example in the USA [
], resulting in a much higher proportion of younger individuals who are undiagnosed. In addition, older individuals are more likely to have developed complications, leading to diagnosis, thus reducing UDM proportion in older age groups. Currently, the estimates only use a total proportion of UDM applied across all age groups and this may underestimate the proportion for younger age groups and over estimate for older ones.
Fig. 3Age-specific proportions (%) of undiagnosed diabetes mellitus (UDM) in adults (20–79 years), 2013. Age-specific UDM proportions were given only in a small number of data sources. One set of age-specific data from each IDF Region is presented [
There is also evidence from nationally representative studies of variation in the proportion of UDM between urban and rural settings. More studies reporting nationally representative figures that take this variation into account would benefit the precision of these estimates.
6. Conclusion
The results of this global study confirm the alarmingly high proportions of UDM in many areas of the world. Undiagnosed diabetes is harmful and costly; both financially and in terms of complications for individuals, communities, and health systems. Nonetheless, it is imperative that the response to these data should be appropriate to the varying capacities of national health systems. Even in countries with the most developed health systems, the proportion of patients not achieving target measurements are disturbingly high; in the UK, over 35% of patients with diabetes do not attain their target HbA1c, and almost 50% do not reach their target blood pressure [
]. The provision of effective care to people with diabetes must be the first priority, before considering strategies to improve screening for UDM and thus identifying more people with diabetes.
The finding that almost half of all diabetes still remains undiagnosed suggests that the current levels of awareness are alarmingly low. Lack of diagnosis of diabetes is a grave burden on health and financial systems throughout the world, and proportions are highest in developing countries. Addressing this challenge will require gradual and careful work, beginning with the strengthening of health systems in preparation for effectively accommodating and treating a higher number of people with diabetes diagnosed by subsequent screening programmes. In addition, more high-quality studies are needed to further understand the burden of UDM and allocate appropriate resources to closing the gap in diabetes care, and the authors hope that the transparency of this methodology will encourage the involvement of the research community.
Conflict of interest
The authors declare that they have no potential conflict of interest, including specific financial interests, relevant to the subject of this manuscript.
Funding
The 6th edition of the IDF Diabetes Atlas was supported by the following sponsors: Lilly Diabetes, Merck and Co, Inc., Novo Nordisk A/S supported through an unrestricted grant by the Novo Nordisk Changing Diabetes® initiative, Pfizer, Inc., and Sanofi Diabetes.
Acknowledgements
With many thanks to Dr Lydia Makaroff for carefully reviewing this manuscript, and to the IDF Diabetes Atlas Committee, and in particular Dr David Whiting, Professor Jonathan Shaw and Professor Ian Hambleton for their assistance and expertise in developing the estimates.
Appendix A. Supplementary data
The following are Supplementary data to this article:
Estimates of direct medical costs for microvascular and macrovascular complications resulting from type 2 diabetes mellitus in the United States in 2000.
United Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects: The 2012 Revision [Internet]. New York; 2013. Available from: http://esa.un.org/wpp/Documentation/publications.htm.
00384-7&id=gr1.jpg){kind=link}
00384-7&id=gr2.jpg){kind=link}
00384-7&id=gr3.jpg){kind=link}